Definition of three minimal deleted regions by comprehensive allelotyping and mutational screening of FHIT,p16INK4A, and p19ARF genes in nasopharyngeal carcinoma




Recurrent deletion on a chromosomal location in tumor cells can be detected by frequent allelic loss and generally is considered to be an indication of the existence of a tumor suppressor gene (TSG) in the region. In the current study, using fluorescent-labeled, high-density microsatellite markers for allelotyping, the authors pinpointed three minimal deleted regions (MDRs) and screened mutations of putative TSGs on chromosomes 3, 9, and 11 in nasopharyngeal carcinoma (NPC) cases occurring in Taiwan.


A total of 133 informative microsatellite markers were used on chromosomes 3, 9, and 11 with an average marker density of 4 centimorgans (cM) for the allelotyping of genomic DNAs isolated from NPC tissues and their corresponding lymphocytes in 48 patients. The correlation between allelic loss and the clinicopathologic parameters of NPC tissues was examined. In addition, putative TSGs including FHIT, p16INK4a, and p19ARF were selected for mutation screening to investigate their potential participation in NPC tumorigenesis.


Of 3787 informative allelotyping data, 25 frequent allelic losses (or loss of heterozygosity [LOH]) in 13 cytogenetic loci were identified based on a deletion frequency that was greater than the average of allelic loss on that particular chromosome. Several significant associations were determined after statistical analysis of the correlation between allelic loss and clinicopathologic parameters. The allelic losses by D9S318 and D11S1304 were associated with N2/N3 (P = 0.035 and P = 0.005, respectively), and those by D9S905 and D11S1304 were associated with grouped American Joint Committee on Cancer (AJCC) Stage III/IV samples (P = 0.022 and P = 0.017, respectively) of NPC tissues. In addition, three MDRs were revealed on 3p25.3-24.1 (< 19 cM), 3p23-21.31 (< 9 cM), and 11q22.1-23.2 (< 8 cM). To examine somatic mutations in previously reported TSGs located near these frequent LOH loci, three candidate genes, p16INK4a, p19ARF, and FHIT, were analyzed. Point mutations in the coding region of FHIT and in the intron 1 splicing acceptor site of both p16INK4a and p19ARF were detected in NPC cell lines. Sequence analysis of both p16INK4a and p19ARF transcripts revealed that the point mutation resulted in skipping of exon 2 and the generation of shorter transcripts.


High-density allelotyping permitted the discovery of 3 MDRs on 3p25.3-24.1 (< 19 cM), 3p23-21.31 (< 9 cM), and 11q22.1-23.2 (< 8 cM) and a correlation was determined between allelic loss and clinicopathologic parameters of NPC tissues. More important, one somatic mutation in NPC cell lines on the intron 1/exon 2 splicing acceptor site of the INK4a/ARF locus was found to result in exon 2 skipping both p16INK4a and p19ARF transcripts, which presumably inactivates the functions of both the p16INK4a and p19ARF proteins. Cancer 2002;94:1987–96. © 2002 American Cancer Society.

DOI 10.1002/cncr.10406

Nasopharyngeal carcinoma (NPC) is one of the most common head and neck malignancies in Southeast Asia and is believed to have a multifactorial etiology.1, 2 The reported NPC incidence of 100,000 persons per year generally is < 1 person in the entire world, but this has increased to 5.12 persons in Taiwan and 12.9 persons in Southern China.2, 3 Both environmental and genetic factors have been implicated in the tumor formation of NPC. Among environmental factors, the consumption of Cantonese salted food and Epstein–Barr virus (EBV) infection have been documented most often.4 Using in situ hybridization, EBV DNA has been reported to be detected in 81.7% and 100%, respectively, of NPC tissues with biotin-labeled and radioisotope-labeled probes.5 In addition to environmental factors, familial clustering of NPC also has been reported.6 For example, Chinese NPC patients have higher frequencies of human leukocyte antigens (HLA) A2 and BW46.7 A linkage study based on affected sib pairs suggested that a gene closely linked to the HLA locus confers a greatly increased risk of NPC.8 In Taiwan, the curative 5-year survival rate is > 60% in all NPC patients and 80% in early-stage NPC patients after radiotherapy and/or chemotherapy.9 Unfortunately, the incidence of NPC peaks at a productive age of 45–55 years. Therefore, early diagnosis developed from molecular genetic studies of NPC formation will potentially improve treatment and clinical management.

Molecular aberrations on cancer chromosomes have been shown to be associated with the development of solid tumors.10 The molecular mechanisms of chromosomal aberrations indicate that the process of tumorigenesis is the accumulation of chromosomal damages such as the activation of oncogenes, the inactivation of tumor suppressor genes (TSGs), and the mutation of mutator genes that are engaged in the repair, replication, and stability of a genome. These chromosomal alterations, especially loss of heterozygosity (LOH), have been applied successfully to localize the regions of recurrent deletions and ultimately facilitate the positional cloning of cancer genes.11 Indeed, many of the TSGs, including DCC and APC in colon, DPC4 in pancreatic carcinoma, and pTEN in prostate carcinoma and other malignancies, initially were localized through evidence of LOH and eventually identified using the positional cloning technique.11–13 Previous cytogenetic studies of NPC biopsy specimens and cell lines have detected deletions of the short arms in chromosomes 3 and 9. LOH was found on chromosome arms 3p, 9p, and 11q in NPC using restriction fragment length polymorphism, and more recently by the use of microsatellite markers.14–17 Recent studies of genome-wide allelotyping by 382 microsatellite markers with 10-centimorgan (cM) resolution on 27 microdissected primary NPC tissues also detected high frequencies of allelic imbalance on chromosome arms 3p, 9p, 9q, 11q, 12q, 13q, 14q, and 16q.18 Because recurrent deletions were found to occur mainly on chromosomes 3, 9, and 11 in NPC, it is important to refine these deletions further for positional candidate cloning of TSGs and to develop molecular genetic methods for better treatment and management of NPC.


Patients, Specimens and Cell Lines

Fresh biopsy specimens from 30 primary tumors and surgical tissues from 18 patients with recurrent NPC (13 cases at the nasopharynx and 5 at the neck) were obtained from the Department of Otolaryngology, National Taiwan University Hospital. Because lymphoid infiltration has been a general concern in NPC tumor samples, these NPC tissues were resected carefully and confirmed by pathology. The corresponding blood lymphocyte also was collected as a control. Genomic DNAs were extracted and purified with phenol-chloroform extraction, followed by ethanol precipitation. The recurrent intervals between the development of primary and recurrent NPC cases ranged from 11–84 months with the median of 51 months. All the NPC cell lines in the study were cultured in 10% fetal bovine serum in Dulbecco modified Eagle medium (DMEM). HONE-1 was a gift from Dr. Ching-Hwa Tsai at the Department of Bacteriology and Institute of Microbiology, School of Medicine, National Taiwan University.19 CNE-1 and CNE-2 were obtained from Professor Yong-Sheng Zong at the Sun Yat-Sen University of Medical Science, People's Republic of China.20, 21

Allelotyping by Microsatellite Markers

A total of 133 informative microsatellite polymorphic markers including 50 dinucleotide, 9 trinucleotide, and 74 tetranucleotide markers spanning chromosomes 3, 9, and 11 were used for allelotyping. The average interval of these markers was approximately 4 cM and only 7 marker intervals were within the range of 10–15 cM. The locations of dinucleotide markers, purchased from PE Applied Biosystems (ABI PRISM Linkage Mapping Set, Version 1, available from URL: (Foster City, CA), are based on the 1996 Genethon map. The locations of trinucleotide and tetranucleotide markers, purchased from the set of Multi-Colored Fluorescent Human MapPairs Markers (Version 8) of Research Genetics (Huntsville, AL; available from URL:, are based on the comprehensive human genetic map of the Center for Medical Genetics at Marshfield, WI. These polymorphic microsatellite markers were labeled with one of the three different fluorescent dyes (FAM, HEX, and TET) on one polymerase chain reaction (PCR) primer to allow a simultaneous analysis of many markers in a lane of the DNA sequencing gel based on the differences in fragment sizes and fluorescent labels. This technique provides a powerful tool with which to interpret allelic loss on a quantitative basis efficiently. PCR reactions were conducted in a volume of 10 μL using 20 ng of genomic DNA, 0.05 μM of fluorescent-labeled and unlabeled primers, 250 μM of each dNTPs, 2.5 mM MgCl2, and 0.5 unit of AmpliTaq DNA polymerase (PE Applied Biosystems, Foster City, CA) with protocols provided by the manufacturers. PCR products with different fluorescent labels and fragment sizes were pooled and mixed with internal fluorescent-labeled (TAMRA) molecular weight markers for subsequent electrophoresis using a PE 377 automated fluorescent DNA sequencer (PE Applied Biosystems). Allele sizing was determined using softwares of GeneScan Analysis version 3.0 and Genotyper version 2.5. We used both TAMRA-labeled GS-500 size standard and parents' genomic DNAs from CEPH families 1331 (Quantum Biotechnologies Inc., Montreal, Quebec, Canada) as fragment calling standards. For each gel image, we employed two persons for lane tracking and allele sizing of markers independently. For each marker, the allelic loss ratio was calculated by the formula of (T1/T2)/(N1/N2), in which T1 and T2 are the values of two peaks derived from tumor and N1 and N2 are from normal tissue. In the current study, we utilized a more stringent definition for the presence of an LOH locus. First, peak values of both allele areas and allele heights were used to calculate allelic loss ratio and thus two sets of data were generated for each marker. Second, only when both sets of data were greater than 2-fold, instead of the 1.5-fold commonly used by other studies, the marker region was considered as an LOH locus.

The cytogenetic locations of microsatellite markers are based on the Genome Database (GDB) (Available from URL: The recombination frequency of markers measured with cM is based on the comprehensive human genetic map of the Center for Medical Genetics at Marshfield, WI.22 For determining the overlapping minimal deleted regions (MDRs) the markers were listed and grouped due to continuation of frequent LOH regions. The candidate TSGs located near the cytogenetic location are based on The Tumor Suppressor and Oncogene Directory of the Cancer Genome Anatomy Project (CGAP) database at the National Cancer Institute (Available from URL:

Statistical Analysis

To examine the association between LOH and the TNM clinical classification of NPC tissues,23 we established a cross-tabulation showing the number of cases in each classification with or without LOH. Because the majority of numbers in each classification are small after allelotyping, we reclassified the categories into 2 categories for each classification. For classification of the primary tumor, we pooled T1/T2 for one category and T3/T4 for another category. For classification of regional lymph nodes, we pooled N0/N1 for one category and N2/N3 for another category. For stage grouping of the nasopharynx in the TNM system, we pooled Stages I/II for one category and Stages III/IV for another category. We only included markers with allelic loss frequencies higher than the 75th percentile of that in all studied markers along that chromosome for statistical analysis. We used the two-tailed Fisher exact test to examine the null hypothesis of no association between LOH with the reclassified TNM classification at those specific loci. All the analyses were performed using PROC FREQ in the SAS software, Version 6.12 (SAS Institute, Inc., Cary, NC). A P < 0.05 was considered to be statistically significant.

Mutational Analysis of TSGs

For sequence analysis of the coding region of FHIT, exon specific primers for exons 5–9 and nested PCR primer pairs were used for genomic DNAs isolated from tissues and for cDNAs prepared from cell lines, respectively.24 For sequence analysis of p16INK4a and p19ARF genes in both tissues and cell lines, exon specific primers were selected based on a previous report25 except the exon 1β of the p19ARF gene with primer pairs of Ebf: TCCCAGTCTGCAGTTAAGG and Ebr1: GTCTAAGTCGTTGTAACCCG. The primer pairs for reverse-transcriptase-PCR (RT-PCR) products of mutant and wild-type p16INK4a and p19ARF transcripts are p161A (CGGAGAGGGGGAGAGCAG) plus p19arf (CCTGTAGGACCTTCGGTGAC) and p19arf-F (GGTTTTCGTGGTTCACATCC) plus p19arf, respectively. Two primers (DADEX1: GCAGTTATGTCGGCGTCGGTAG and DADEX2: GTTCTGTGGGTTGATCTGTATTC), for the DAD 1 gene were used as a quantitative control for RT-PCR in all cell lines.


Allelotyping by Microsatellite Markers

Based on 3787 informative data, we detected 78 (59%) informative loci per sample and 70% (34/48) fractional allelic loss (FAL) on chromosomes 3, 9, and 11 in NPC samples. The frequencies of FAL for each chromosome arm were 96% (46 of 48 cases) for 3p, 85% (41 of 48 cases) for 3q, 42% (20 of 48 cases) for 9p, 71% (34 of 48 cases) for 9q, 38% (18 of 48 cases) for 11p, and 91% (44 of 48 cases) for 11q. Figure 1 shows examples of allelic loss on cancer chromosomes detected by four markers on allelotyping analysis. The average allelic loss frequencies for all markers on NPC chromosomes 3, 9, and 11 were 24%, 20%, and 21%, respectively. To exclude the random event of allelic loss and to avoid the deletion-masking problem resulting from normal cells contaminated in the tumor samples, we only considered loci in which the number of allelic losses was > 10 cases and the frequency of allelic loss was above the average for that particular chromosome. Table 1 lists qualified markers under stringent criteria described earlier as well as their cytogenetic locations and surrounding candidate TSGs. We identified 13 common LOH loci and regions commonly observed on chromosomes 3, 9, and 11 in NPC samples. Among them, 8 frequent LOH loci at 3p25.2-25.1, 3q22.1-23, 3q26.2-27.1, 3q29, 9q21.32-22.2, 9q34.12-34.3, 11p15.5-15.4, and 11q24.3-25 were identified for the first time. In addition, four previously reported loci also were identified including 3p21-14, 9p21-23, 11q13-21, and 11q21-23. Among these, 11q13-21 and 11q21-23 were refined to 11q14.3 and 11q22.1-23.2, respectively, in the current study, whereas the 3p21-14 locus was found to split into 3p24.3-p21.31 and 3p14.3-13 due to interruption of informative markers (data not shown).14–17

Figure 1.

The representative allelotyping analysis of trinucleotide (D3S2418 and D9S2157) and tetranucleotide (D3S2406 and D11S4464) markers on genomic DNAs isolated from different nasopharyngeal carcinoma tissues and corresponding lymphocytes using Genotyper analysis software (PE Applied Biosystems). The electropherogram plot of the top row represents the tumor alleles in comparison with the corresponding normal alleles in the bottom row. The table under the electropherogram plot indicates the name of the marker, the allelic ratio of the peak height, and the allelic ratio of the peak area.

Table 1. High Frequency of Allelic Loss Regions on Chromosome 3, 9, and 11 of NPC Tissues
Cytogenetic locicMMarkersFrequenciesCandidate TSGs
  1. NPC: nasopharyngeal carcinoma; cM: centimorgans; TSGs: tumor suppressor genes.

3p25.2-25.126.25D3S454526% (11/42)
3p24.3-p21.3138.28ATA9B0965% (20/31)VHL, MLH1, TGFBR2
41.56D3S151638% (14/37)CTNNB
49.18D3S232761% (22/36) 
55.11D3S453530% (13/44) 
57.92D3S243228% (10/36) 
61.52D3S176831% (11/35) 
62.05D3S241146% (18/39) 
67.94D3S176750% (14/28) 
3p14.3-13102.64D3S240628% (13/46)PTPRG, FHIT
103.72D3S303933% (11/33) 
3q22.1-23143.94D3S231634% (10/29) 
146.60D3S167142% (14/33) 
3q26.2-27.1181.87D3S152533% (14/43) 
188.29D3S304133% (15/45) 
3q29224.88D3S230646% (11/24)NFKB2
9p22.1-13.351.21D9S76126% (11/43)p16INK4a, p19ARF, p15
9q21.32-22.296.46D9S31832% (12/38)PTC
9q34.12-34.3163.84D9S90536% (16/45)TSC1
11p15.5-15.48.97D11S236252% (16/31)p57KIP2
11q14.387.88D11S201526% (12/46) 
11q22.1-23.2101.75D11S201723% (11/47)ATM, PPP2R1B, ST3
105.74D11S198638% (17/45) 
108.59D11S96523% (10/43) 
11q24.3–25141.91D11S130444% (15/34)BARX2

Correlation of LOH Loci with Clinicopathologic Parameters

Table 2 shows the results of statistical analysis of the association between allelic loss loci and the TNM parameters of NPC tissues. Two allelic loss loci, 9q21.3-22.2 and 11q24.3-25, detected by D9S318 (P = 0.035) and D11S1304 (P = 0.005), respectively, were observed frequently in NPC tissues classified at N2/N3. Two other allelic loss loci, 9q34.1-34.3 and 11q24.3-25, detected by D9S905 (P = 0.022) and D11S1304 (P = 0.017) respectively, were associated with grouped AJCC Stages III/IV. Recent studies have indicated that the allelic loss loci on 9q22-23 and 11q24.3-25 are associated significantly with advanced stage disease in ovarian, breast, and cervical carcinomas.26–30

Table 2. Association between LOH and Tumor Classification Systems
LociMarkersCategoryNo. of LOHNo. of no LOHP value
  1. LOH: loss of heterozygosity.

9q21.3-22.2D9S318Regional lymph nodes classification
N0 or N13 (25%)17 (65%)0.035
N2 or N39 (75%)9 (35%)
9q34.1-34.3D9S905Stage grouping in TNM system
I or II2 (13%)13 (46%)0.022
III or IV14 (87%)15 (54%)
11q24.3-25D11S1304Regional lymph nodes classification
N0 or N14 (27%)15 (79%)0.005
N2 or N311 (73%)4 (21%)
Stage grouping in TNM system
I or II3 (20%)12 (63%)0.017
III or IV12 (80%)7 (37%)

Determination of MDRs

After comprehensive allelotyping on NPC chromosomes 3, 9, and 11, the allelic loss pattern of profile of each NPC case was determined based on the relative genetic distances between markers. To facilitate potential positional cloning of TSGs, the interstitial deletions in each NPC sample were aligned to determine minimal LOH overlapping regions on continuous and frequent LOH markers across NPC cases. In Figure 2, we selected 2 continuous allelic loss regions from informative cases in Table 1 (3p and 11q) and defined three distinct LOH overlapping regions on 3p25.3-24.1 (D3S2403-D3S4535, < 19 cM), 3p23-21.31 (D3S1768-D3S1766, < 9 cM), and 11q22.1-23.2 (D11S2000-D11S965, < 8 cM) from current study data. Deletion of each MDRs was found in 75% (36 of 48 cases), 65% (31 of 48 cases), and 67% (32 of 48 cases) of NPC tumors to loci 3p25.3-24.1, 3p23-21.31, and 11q22.1-23.2, respectively. Comparison of our MDRs with results previously generated from LOH analysis or from functional suppression of NPC tumorigenicity by other groups indicated that the three MDRs we determined currently are the smallest consensus-deleted regions found on NPC chromosomal regions.14–18

Figure 2.

The schematic patterns of minimal deleted regions (MDRs) in selected nasopharyngeal carcinoma (NPC) tissues with frequent loss of heterozygosity (LOH). The markers located in frequent LOH regions with larger letter also were illustrated in Table 1. The dark, gray, and open boxes represent markers with LOH, uninformative markers, and retention of heterozygosity, respectively. The number on the top of each column represents the case number. The aligned black lines along the boxes on the right indicate the consensus regions assigned by our studies (a, c, and g) and previous reports in NPC (b, e, f, h, and i).14–18, 42 The dark line “d” represents the critical region assigned by functional suppression of tumorigenicity in an NPC cell line.42 The dark lines “f” and “i” represent critical deletion regions assigned by LOH scan on microdissected NPC samples.18

Mutation Analysis of Three Candidate TSGs: FHIT, p16INK4a,p19ARF

Having identified frequent LOH regions on chromosome 3, 9, and 11, we next searched for TSGs residing in these regions and investigated their possible involvement in the tumorigenesis of NPC. Three candidate genes, FHIT, p16INK4a, and p19ARF, located in two cytogenetic loci (3p14.3-13 and 9p22.1-13.3), were selected for mutation analysis. The FHIT gene was identified and cloned in a hereditary renal carcinoma-associated t(3;8) translocation.31 Since that time, the deletion of its genomic structure, aberrant transcription of the FHIT gene, and the reduction or absence of FHIT protein frequently was found in many tumors.32 Recently, the heterozygous FHIT mice (+/−) were reported to develop multiple tumors after carcinogen challenge.33 These findings strongly suggest the function of FHIT as a TSG. In addition, several lines of evidence supported findings that both p16INK4a and p19ARF genes function as TSGs in various malignancies. p16INK4a functions as an inhibitor of cdk4/6 and blocks the passage from G1 to S-phase with functional pRB in cells.34 p19ARF induces both cell-cycle arrest and apoptosis, and blocks myc/ras transformation in a p53-dependent manner. Homozygote deletion and aberrant transcriptional inactivation due to hypermethylation of p16INK4a were detected in many types of tumor cell lines including NPC.35–38 Because no point mutation of p16INK4a in the coding regions was reported and because to our knowledge the role of p19ARF is unknown for the dual transcripts in NPC, we performed mutational screening on the INK4a/ARF locus to examine their potential participation in the tumorigenesis of NPC.

Although frequent LOH has been observed in these cytogenetic loci, no mutation was detected in the coding region of genes FHIT, p16INK4a, and p19ARF from 21 NPC tissues (data not shown). However, in three available NPC cell lines, two point mutations, Ser77Pro and Gln9-Arg, were found in the coding region of the FHIT gene in HONE-1 cells, whereas synonymous mutations were detected in CNE-1 cells (Table 3). Although the biologic consequence of the two point mutations found in HONE-1 cells remains unclear, they represent only a few cases of point mutation of the FHIT gene reported to date.32 Furthermore, an A to C transversion was detected on the sharing junction of the intron 1/exon 2 splicing acceptor site of the INK4a/ARF locus in all three NPC cell lines tested (Table 3). The identification of this mutation prompted us to investigate whether the splicing of both p16INK4a and p19ARF transcripts are affected. As shown in Figure 3B, smaller RT-PCR products of both p16INK4a and p19ARF were detected from NPC cell lines compared with the control cell line. Sequencing analysis of aberrant p16INK4a and p19ARF PCR products indicated that the mutated intron 1/exon 2 splicing acceptor site leads to a skipping of exon 2 and a joining of exon 1 to exon 3 in both transcripts (Fig. 3C–G).

Table 3. Alterations of p16INK4a, p19ARF, and FHIT Genes in NPC Cells
NPC cell linesCodonNucleotide changeAberrations
  1. NPC: nasopharyngeal carcinoma.

  2. In1/Ex2 acceptor site represents for the acceptor site of the intron 1/exon 2 splicing junction.

FHIT gene
 HONE-177TCT → CCTSer → Pro
90CAG → CGGGln → Arg
 CNE-182CAG → CAANone
p16INK4a and p19ARF genes
 HONE-1In1/Ex2 AcceptoragGT → cgGTSplicing defect
 CNE-1In1/Ex2 AcceptoragGT → cgGTSplicing defect
 CNE-2In1/Ex2 AcceptoragGT → cgGTSplicing defect
Figure 3.

Splicing defect of the INK4a/ARF locus in nasopharyngeal carcinoma (NPC) cell lines. (A) The genomic structure of p16INK4a and p19ARF genes. The “∗” indicated the mutation site of the exon 2 acceptor. The gray, black, and empty boxes indicate the coding exons of p19ARF, the coding exons of p16INK4a, and the untranslated regions of both transcripts, respectively. The dashed and dotted lines indicate the prospective splicing/joining of exons in transcripts of p16INK4a ΔE2 and p19ARF ΔE2. (B) The electrophoresis of reverse transcriptase-polymerase chain reaction (RT-PCR) products for the transcripts of the INK4a/ARF locus in NPC cell lines. Lanes 1–4 represent the RT-PCR products from cell lines HONE-1, CNE-1, CNE-2, and Hep3B (a liver carcinoma cell line used as a positive control) of both genes. The sequencing chromatograms at the junction of exons from the wild-type (Hep3B) and mutant (HONE-1) INK4a/ARF locus including exon 1β-exon 2 (C), exon 1β-exon 3 (D), exon 1α-exon 2 (E), exon 2-exon 3 (F), and exon 1α-exon 3 (G).


In the current study, we performed high-density screening of LOH loci using microsatellite markers to dissect and refine the common LOH loci located on chromosomes 3, 9, and 11 in NPC tissues. To circumvent the masking effect from contaminated normal cells and to ensure detection of frequent LOH loci resulting from tumorigenesis, stringent criteria were used to define LOH in the current study. As indicated in Figure 1 and Table 1, the low background peaks in the tumor samples and the discovery of new allelic loss loci suggested limited contamination of nonneoplastic cells in our tumor specimens that could mask LOH detection. Indeed, LOH frequencies on the chromosome arms detected in the current study are similar to results obtained from microdissected NPC samples reported previously.18 The newly revealed allelic loss loci, with frequencies ranging from 26–52%, suggest that high-density marker screening might disclose subtle recurrent deletions on NPC chromosomes. In addition, MDR analysis allowed us to detect and refine the three currently smallest MDRs in NPC on 3p25.3-24.1, 3p23-21.23, and 11q22.1-23.2. The observation of several common LOH loci on the same chromosome is consistent with previous reports regarding a number of tumors.39–41 The presence of putative TSGs on each of chromosomes 3, 9, and 11 is supported by the suppression of tumorigenicity after chromosome transfer into various carcinoma cells including NPC cells.42–46

We found certain LOH loci are associated significantly with late-stage NPC development such as the allelic losses detected by D9S318 and D11S1304, which were shown to be associated with N2/N3 classifications, and by D9S905 and D11S1304 with grouped AJCC Stages III/IV samples of NPC tissues. These associations imply that putative TSGs residing in these deleted regions are involved in the development of NPC into advanced stages, such as mass growth, lymph node invasion, and/or metastasis. The allelic loss detected by D9S318 in the cytogenetic location 9q22 also was reported previously to be associated with the presence of lymph node metastasis of primary breast carcinoma26 and with late events in the development of squamous cell carcinoma of the skin.47 The 9q22.3 region contains a known TSG, the human homolog of Drosophila' patched' (ptc) gene, PTC/PTCH. The PTC gene encodes a transmembrane protein that represses transcription in specific cells of genes encoding members of the TGF-β and Wnt families of signaling proteins.48 Although reduction of cell-cell adhesion through signaling proteins in Wnt families is essential for tumor invasion, to our knowledge the potential association between PTC/PTCH function and metastasis or late-stage tumorigenesis in NPC remains to be determined. Another allelic loss region on 9q34, detected by D9S905, also was frequently deleted in other sporadic tumors including bladder and squamous cell carcinoma.49 It is interesting to note that the association of allelic loss in 9q34 with late-stage NPC also was observed in a previous study of allelic loss of sporadic primary NPC cases from Southern China.50 As indicated in Table 1, the candidate TSG in 9q34 is the TSC1 gene-identified gene mutant in tuberous sclerosis-1.51 Overexpression of the TSC1 gene in rat fibroblasts inhibits growth via disregulation of the cell cycle and acts as a tumor suppressor.52 Because the function of TSC1 involvement in the late stages of malignancy remains unclear, further functional studies of TSC1 and other potential TSGs on 9q34 should be performed to better understand the late-stage progression of NPC.

Several lines of evidence have indicated that the 11q24-25 region may contain a late-acting TSG that is prognostically significant for ovarian carcinoma, breast carcinoma, cervical carcinoma, malignant melanoma, and other malignancies.53 The findings in the current study of an association between allelic loss in 11q24-25 and advanced stages of NPC further support the theory of putative TSG participation in the disease progression of multiple tumors. Recent cloning of a putative TSG located in 11q24-25 suggested that the human homeobox gene BARX2 induces cadherin 6 expression and acts as a functional suppressor of ovarian carcinoma progression.54 The gene expression of human BARX2 and cadherin 6 was shown to be at a significantly lower level in both ovarian cell lines and tissues. Transfection of BARX2 into ovarian cell lines without an endogenous gene (OAW42) inhibited Matrigel (see for a detailed explanation about Matrigel) invasion, haptotactic cellular migration to a Type IV collagen signal, and adhesion to Type IV collagen-coated plates. The BARX2 is a transcriptional factor and is homologous with the Drosophila Bar class of homeobox-domain-containing proteins. The human BARX2 transcriptional factor regulates transcriptions of target genes such as specific CAMs and calcitonin genes for tumor progression and adverse survival.54, 55 It will be interesting to exam the down-regulation of BARX2 and cadherin 6 in the tumor progression of other major malignancies, including NPC.

Although no mutation was found in p16INK4a, p19ARF, and FHIT from NPC tissues, we did detect mutations on each of the three genes in the NPC cell lines. It generally is believed that the genotype and phenotype of a cancer cell line represents a snapshot of the tumor at the time the biopsy was taken. Because the genetic changes required for cell immortalization mainly are late events in disease progression and because the majority of primary tumors are not immortal, it is not surprising that we detected mutations in the NPC cell lines but not in primary NPC tissues.56 It is interesting to note that recent studies of both p16INK4a and p19ARF function suggested that they act in overlapping pathways in cellular immortalization.57 Nevertheless, the rare point mutations (Ser 77 Pro and Gln 90 Arg) of the FHIT gene were detected in HONE-1 cells and a point mutation at the intron 1/exon 2 splicing acceptor site of the INK4a/ARF locus was detected in all three NPC cell lines. This INK4a/ARF point mutation was reported previously in two NPC cell lines (CNE-1 and CNE-2) and a tumor cell line (HTB-94) derived from a chondrosarcoma, suggesting that this site is a mutational hotspot.37, 58 To determine whether this mutation could affect both p16INK4a and p19ARF transcripts, we performed RT-PCR, DNA sequencing, and Western blot analysis to examine the molecular consequences of the mutation. Our sequencing analysis indicated that the mutation results in exon 2 skipping both p16INK4a and p19ARF transcripts, which presumably inactivates the functions of both the p16INK4a and p19ARF proteins. The transcript resulted from the linkage of exon-1α to the downstream available exon 3 is predicted to encode a novel protein that contains the first 50 amino acids of p16INK4a and an additional 39 amino acids from exon 3 with a different reading frame to p16INK4a. Another transcript derived from a joint of exon-1β and exon 3 presumably encodes a protein with the first 64 amino acids from p19ARF and the last 4 amino acids from the c-terminal of p16INK4a. However, Western blot analysis with antihuman p19ARF and p16INK4a antibodies detected positive signals from control HeLa cells but failed to detect any signal from lysate of these NPC cell lines (data not shown). Whether these two mutated proteins are untranslated, unstable, or unable to be detected by the antibodies used due to the extensive sequence aberration remains to be determined.


The authors thank Chia-Wen Liang, Tai-Tsung Chen, Lian-Du Wang, Chuan-Chuan Chao, The-Li Su, Tzung-Shiahn Sheen, and Ruey-Hwa Chen for their valuable technical assistance and advice.